Household Energy Consumption Forecasting Using LSTM and ARIMA Models: A Comprehensive Analysis

Accurately forecasting residential energy consumption is essential for efficient energy management and optimization. This study aims to develop predictive models to forecast household energy consumption for the next 15 days (multi-step ahead forecasting) of a given month, based on the previous 15 days of data. Using minute-level data collected over 15 days from a residential setup with an ESP32 microcontroller and a PZEM-004T energy monitoring module, this study examines the capabilities of ARIMA and LSTM models in predicting future energy consumption. The data underwent extensive preprocessing, including normalization, differencing, and stationarity tests, to ensure suitability for modeling. The ARIMA model, configured with an order of (1, 2, 15), and the LSTM model, featuring four LSTM layers with varying dropout rates, were evaluated for their forecasting accuracy. The ARIMA model achieved a Root Mean Squared Error (RMSE) of 0.0217 and a Mean Squared Error (MSE) of 0.0009. Conversely, the LSTM model demonstrated superior performance with an RMSE of 0.0196, a test loss of 0.000851, and a lower Mean Absolute Error (MAE) of 0.015, compared to the ARIMA model's MAE of 0.025. Additionally, the LSTM model achieved a more favorable Mean Absolute Percentage Error (MAPE) of 3.0%. These findings highlight the effectiveness of LSTM models in capturing complex temporal patterns and improving forecasting accuracy compared to traditional ARIMA models. This research provides valuable insights into advanced energy forecasting methodologies, laying the groundwork for future work in residential energy optimization.\section*{\small Article Highlights}\begin{itemize}\item LSTM models outperform ARIMA in predicting household energy usage by capturing complex patterns.\item Minute-level data enhances the accuracy of short-term energy consumption forecasting.\item Real-time LSTM-based predictions could optimize energy management in smart homes.\end{itemize}